How the score works
The AI Missed Opportunity Score is a free AI readiness check for small businesses. It reads your public website and scores, from 0 to 100, how your business handles the work competitors already hand to AI. This page is the full method: what it checks, how each check counts, and when it refuses to give a score.
Get my free scoreAt a glance
- Scale0 to 100, in four bands: Behind, Catching up, Solid, Ahead.
- AreasFour, weighted: lead capture and response 30%, customer service 20%, content and social 25% and data and foundations 25%.
- Checks31 checks: 28 carry points and 3 add context only. Checks that do not fit your type of business are left out.
- What it readsYour home page and up to 5 more pages, the scripts your home page loads, both web addresses (with and without www) and Google PageSpeed on mobile. Only what any visitor can see.
- Your answersUp to 5 yes or no questions for things a public website cannot show.
- Evidence ruleNo score unless at least 5 scored checks could be assessed.
- Fact-checkEvery gap is checked by a second pass against the pages read before you see it.
- CostFree. It needs your website address and an email.
The four areas and their weights
Each area covers one kind of work that businesses using AI now do faster or around the clock. The weights say how much each area matters to a small business, so the area closest to revenue counts most.
| Area | What it asks | Weight |
|---|---|---|
| Lead capture and response | Can a visitor reach you, get an answer and book, at any hour, without waiting? | 30% |
| Customer service | Can customers find answers and help themselves without writing to you? | 20% |
| Content and social | Does the site show proof and recent activity, and link to where you publish? | 25% |
| Data and foundations | Can you measure what works, keep every enquiry, and does the site load fast? | 25% |
The 31 checks
Every check is detected by code from the pages read, never guessed by a model. Each one ends in one of three states: it earns its points, it does not, or it could not be assessed. A check that could not be assessed is left out of the sums. It never counts against you.
Lead capture and response
9 checks · 30% of the score| Check | Earns points when | Points |
|---|---|---|
| Live chat on the site | Found | 8 |
| AI answering your chatIf the site cannot show it, you are asked. | Found | 4 |
| Online booking | Found | 8 |
| Contact form is the only way in | Not the case | 6 |
| Slow reply promised on the site | Not the case | 4 |
| Phone, WhatsApp or email only | Not the case | 6 |
| Captures visitors who are not ready to enquireIf the site cannot show it, you are asked. | Found | 4 |
| Quotes requested by handContext only: no points, but it can raise a fix. Only for trades and home services, professional services, manufacturing and wholesale, real estate or agencies and creative. | No points | 0 |
| Opening hours with no after-hours pathOnly for trades and home services, health and clinics, fitness and wellness, beauty and salons, restaurants and hospitality, retail and e-commerce, real estate or professional services. | Not the case | 4 |
Customer service
6 checks · 20% of the score| Check | Earns points when | Points |
|---|---|---|
| FAQ or help page | Found | 6 |
| Help centre where customers find answersOnly for retail and e-commerce, software and SaaS, travel and events or restaurants and hospitality. | Found | 3 |
| Support runs through an email inboxOnly for retail and e-commerce, software and SaaS, travel and events or restaurants and hospitality. | Not the case | 5 |
| Customer login, order tracking or portalOnly for retail and e-commerce, software and SaaS, health and clinics, fitness and wellness, education and coaching, travel and events or restaurants and hospitality. | Found | 3 |
| Customer reviews on your site | Found | 3 |
| Written policies and process pagesOnly for retail and e-commerce, restaurants and hospitality, travel and events, fitness and wellness, beauty and salons, health and clinics or education and coaching. | Found | 2 |
Content and social
6 checks · 25% of the score| Check | Earns points when | Points |
|---|---|---|
| Social profiles linked from the site | Found | 5 |
| Blog or news section | Found | 4 |
| Content published in the last 90 days | Found | 6 |
| Video on the site or a linked channel | Found | 3 |
| Testimonials or case studies | Found | 3 |
| Link previews when your site is shared | Found | 2 |
Data and foundations
10 checks · 25% of the score| Check | Earns points when | Points |
|---|---|---|
| Analytics installedIf the site cannot show it, you are asked. | Found | 6 |
| Paid ad tracking installedContext only: no points, but it can raise a fix. | No points | 0 |
| CRM or email platform connected to the siteIf the site cannot show it, you are asked. | Found | 5 |
| Automation platform in useIf the site cannot show it, you are asked. | Found | 2 |
| Website platformContext only: no points. | No points | 0 |
| Fast on mobile | Found | 5 |
| Site looks maintained (current copyright year) | Found | 3 |
| Hiring for work an assistant could take | Not the case | 3 |
| Both web addresses load (with and without www) | Found | 4 |
| Search engines can read your business details | Found | 2 |
Thresholds worth knowing
- Fast on mobile means a Google PageSpeed mobile score of 50 or more and the main content showing within 4 seconds.
- Content published in the last 90 days reads the newest date on your blog or news page. With no blog, it is not assessed rather than marked as missing.
- Site looks maintained passes when the copyright year in the footer is this year or last year.
- Not found only counts as missing when the pages could actually be read. A site built by scripts that returns almost no text and almost no links is marked as not assessed instead.
How the number is calculated
Area score. Within an area, the score is the points earned divided by the points possible, counting only the checks that could be assessed, times 100. An area needs at least two assessed checks to get a score. With fewer, it shows no score rather than a number built on one data point.
Headline score. The headline is the weighted average of the area scores, using the weights above. If an area has no score, the remaining weights are scaled up so they still add to 100%.
Example. A business scoring 50 in lead capture and response, 80 in customer service, 40 in content and social, 60 in data and foundations gets 30% × 50 + 20% × 80 + 25% × 40 + 25% × 60 = 56, which is in the Catching up band.
The same pages always give the same score. The rules are fixed in code. A model only writes the wording afterwards: a line about your business, the one-line summaries and the lead finding. Any number in a summary or the lead finding must appear in the scan data, or the original wording is kept.
Score bands
Checks that do not fit your business are left out
A restaurant does not need a quote calculator and a law firm does not need order tracking. The scan places your business in one of 15 types. Your own pick wins; without one, a short model call reads the home page, and a keyword match is the fallback. You can correct the type after the scan and the score is recalculated.
6 checks only apply to some types (they are marked in the tables above). When the type is unknown, those checks are left out rather than guessed. A check that is left out never scores and never raises a fix.
Sites that have not launched. When the home page reads as a beta, waitlist or coming-soon page, the 14 checks that only matter once customers arrive (such as reviews, booking, logins and support) are set aside and the report says why.
How your website is read
- Pages. The home page first, then up to 5 more pages it links to, at most one of each kind: contact, booking, case studies, services, pricing, FAQ or help, about, blog and careers. Login, sign-up, cart and checkout pages are skipped.
- Readers. Each page is read with Firecrawl first, then a plain request, then Jina Reader if the first two come back thin. A read only counts when it returns a title or more than 300 characters of text.
- Tools behind the page.Analytics, ad pixels, chat widgets and booking tools are often loaded by the site’s own scripts, so the scripts behind the home page are read too.
- Both addresses. The site is opened with and without www, because a visitor who types the broken one sees an error.
- Speed. Google PageSpeed Insights runs a mobile test on the home page.
The scan only sees what any visitor sees. It never logs in, fills in a form or installs anything. A score for the same site is reused for 30 days, unless the scoring engine has changed since.
Questions only you can answer
Some things a public website cannot show. When one of these is not visible, you get a yes or no question instead of a claim that it is missing:
- Does an AI answer the chat on your website?
- Do you collect emails from visitors who are not ready to buy yet (newsletter, free guide, download)?
- Do you track website visitors with analytics (Google Analytics or similar)?
- Do enquiries from your website go straight into a CRM or email platform?
- Do automations (Zapier, Make, n8n or similar) move your leads and data between tools?
Your answers are applied on top of the original scan and the score is recalculated as a new version. Only the questions your scan actually raised can be answered, and only with yes or no.
No score without enough evidence
If fewer than 5 of the scored checks that apply to your business could be assessed, there is no score. You are told the site could not be read well enough, rather than handed a number that is mostly guesswork.
Every gap is fact-checked
Pattern matching misses things a person sees at a glance: a booking page in another language, a video that only appears in the page data, a login drawn by a script. So before a score is saved, a second pass checks every “you are missing this” finding. Claude Sonnet reads the same pages plus the full list of the site’s URLs and gives each gap one of three verdicts:
- Confirmed. Nothing suggests the gap is wrong, so it stands.
- Wrong. The site does have it. The gap flips to a pass, but only if the verdict quotes text that exists word for word in the data. Without that quote it counts as unsure.
- Unsure. There is a hint but no proof, so the check becomes not assessed and stops counting either way.
What you get, and what the paid report adds
Free: the score out of 100, a score and one line for each area, and your fixes ranked by what they likely cost you, each with a do-it-yourself next step. The first fix comes with the AI tools to do it. You see it on screen and by email.
Full report (paid):
- Up to 10 more pages of your site are read, plus your Instagram profile when you have one.
- A Google search finds up to three competitors, and each is checked on seven of the same checks so you see where they are ahead.
- Two more areas: internal operations, from a few questions about what takes up your week, and against your competitors. The six areas are weighted lead capture and response 25%, customer service 15%, content and social 15%, data and foundations 15%, internal operations 15% and against your competitors 15%.
- Every gap with a named tool from a vetted catalogue, what it costs and the first steps.
- A 30-day plan, week by week, biggest payoff first.
The report is checked by code before it is sent: tools can only come from the catalogue, findings that do not hold up are dropped, and numbers that cannot be traced back to the scan are flagged.
Who built it
Felix Mohr, who works with small businesses as their AI business partner. The businesses he works with are in the case files. How your email and the scan are stored is in the privacy policy. Questions about the method: business@mohrwolfe.com.